2족 보행 로봇의 비대칭적인 궤적 생성 알고리즘과 강건한 순응 제어기 설계

Abstract

Joint actuators with a electrical motor, a timing belt, and a harmonic driver system are embedded on each leg of a biped robot. They make the robot's leg relatively heavy, and the center of mass exists far from the hip. The supporting leg also becomes flexible due to the flexibility of harmonic driver systems in each joint and force sensing systems on each foot. The biped robot with such relatively heavy and flexible legs behaves asymmetrically. If the robotic control system requires a lot of works to recover to the normal and symmetric pattern (that is, to eliminate the large error), the recovery works increase more energy consumption for bipedal locomotion. In this work, an asymmetric trajectory generation method for biped robots is proposed to maintain a high level of dynamical postural stability, to increase energy autonomy, and to transit continuous gaits, based on the stability criterion defined in phases. The dynamic postural stability for walking and running motions is defined by the distance of the Zero Moment Point to the boundaries of a stability region and the magnitude of the angular momentum at the mass center. If satisfied with these conditions, the biped robot can walk and run without falling down on the ground. For running motions, the magnitude of the angular momentum at the mass center for flight is determined by the desired motion of the preceding support phase. The support phase is relatively more important to guarantee the dynamic postural stability than the flight phase. First of all, symmetric running and walking patterns are derived from the simplified model with which the essential characteristics of the biped robot are captured according to the gaits. An inverted pendulum model (IP) for walking motion and a spring loaded inverted pendulum model (SLIP) for running motion are used to make the robot behave symmetrically. In general these symmetric motions give us locally optimized trajectories, but does not induce online adaptive motions for gait transitions, velocity change, etc. The online adaptive motions are achieved by the asymmetric IP-like motions and asymmetric SLIP-like motions. The running motion for flight is generated by applying both the linear momentum theorem and the angular momentum theorem to the entire robot as six non-holonomic constraints. The ZMP-Based trajectory modification strategy for support phases is used so that the swing foot and supporting foot do not bounce back in contact with the ground and the soles should be stuck along the contact surface. The biped robot walks and runs frequently contacting with the ground, and also its both legs are switched after contact or support. In order for the biped robot to comply with the ground, a position-based impedance controller is designed with the force modulation algorithm. This controller guarantees a stable landing on the ground and simultaneously tracks the desired trajectories where the desired impedance at the hip link and both legs is specified. For real application, a sliding-mode based impedance controller is proposed as a robust compliance controller such that regulate the position and the contact force simultaneously. This controller is composed of a feedforward control loop and a feedback control loop. A series of computer simulations for two different types of biped robots show that the proposed trajectory and control method satisfy the conditions for dynamic postural stability and make the biped robot more robust to variations in the desired speed, and gait transitions between walking and running. The feasibility of this method has been tested with various experiments on a 12-DOF biped robot without arms. The biped robot could walk on uneven terrain and run successfully with the Froude number $0.142$. The computation time needed for generating feasible trajectories and implementing the impedance control is relatively short, such that they can be applied to the experiments online.